high density 3d-boron nitride & 3d-graphene for high ... · three-dimensional graphene (3d-c),...
TRANSCRIPT
This document is downloaded from DR‑NTU (https://dr.ntu.edu.sg)Nanyang Technological University, Singapore.
High‑Density 3D‑Boron Nitride and 3D‑Graphenefor High‑Performance Nano–Thermal Interface Material
Loeblein, Manuela; Tsang, Siu Hon; Pawlik, Matthieu; Phua, Eric Jian Rong; Yong, Han;Zhang, Xiao Wu; Gan, Chee Lip; Teo, Edwin Hang Tong
2017
Loeblein, M., Tsang, S. H., Pawlik, M., Phua, E. J. R., Yong, H., Zhang, X. W., et al. (2017).High‑Density 3D‑Boron Nitride and 3D‑Graphene for High‑Performance Nano–ThermalInterface Material. ACS Nano, 11(2), 2033‑2044.
https://hdl.handle.net/10356/84059
https://doi.org/10.1021/acsnano.6b08218
© 2017 American Chemical Society. This is the author created version of a work that hasbeen peer reviewed and accepted for publication by ACS Nano, American Chemical Society.It incorporates referee’s comments but changes resulting from the publishing process,such as copyediting, structural formatting, may not be reflected in this document. Thepublished version is available at: [http://dx.doi.org/10.1021/acsnano.6b08218].
Downloaded on 15 Dec 2020 15:22:04 SGT
1
High Density 3D-Boron Nitride & 3D-Graphene for
High Performance Nano-Thermal Interface Material
Manuela Loeblein,†,‡,$ Siu Hon Tsang,|| Matthieu Pawlik,‡ Eric Jian Rong Phua,||, # Han Yong,$
Xiao Wu Zhang,$ Chee Lip Gan||,# and Edwin Hang Tong Teo†,#,*
†School of Electrical and Electronic Engineering, Nanyang Technological University, 50
Nanyang Avenue, Singapore 639798, Singapore
‡CNRS-International NTU Thales Research Alliance (CINTRA) UMI 3288, Research Techno
Plaza, 50 Nanyang Drive, Singapore, Singapore 637553
|| Temasek Laboratories@NTU, 50 Nanyang Avenue, Singapore 639798, Singapore
# School of Materials Science and Engineering, Nanyang Technological University, 50 Nanyang
Avenue, Singapore 639798, Singapore
$Institute of Microelectronics, Agency for Science, Technology and Research (A*Star), 11
Science Park Road, Singapore 117685, Singapore
*Corresponding Author Email: [email protected]
KEYWORDS: three-dimensional graphene, three-dimensional h-BN, nano-thermal interface
material, hot-spot removal, thermal management
2
ABSTRACT: Compression studies on three-dimensional foam-like graphene and h-BN (3D-C
and 3D-BN) revealed their high cross-plane thermal conductivity (62 – 86 Wm-1K-1) and
excellent surface conformity, characteristics essential for thermal management needs.
Comparative studies to state-of-the-art materials and other materials currently under research for
heat dissipation revealed 3D-foam’s improved performance (20 – 30% improved cooling,
temperature decrease by ΔT of 44 – 24°C).
The advancement of semiconductor technology in the era of “more-than-Moore”1 has led to an
increasing challenge in thermal management. Integrated circuits (ICs) are so densely packed that
they heat up within milliseconds,2, 3 resulting in an extreme increase of generated heat4 (e.g.
current power electronics modules can go up to 200 Wcm-2).5 In order to mitigate this problem,
the conventional way of extracting the heat out through the substrate has been improved via
adding heat sinks either on top or below the electronic device. The contact to these heat sinks is
further enhanced through thermal interface materials (TIMs)2 and these TIMs are essential to
reduce the thermal boundary resistance between the chip’s active area and the heat sink.6 Typical
TIMs can be classified into electrically conductive and insulative types. For the former class,
metal alloys containing In, Bi, Sn, Au are the common materials used. Although their thermal
3
conductivities are usually high (i.e. ~50 Wm-1K-1), the use of these TIMs is sometimes limited,
due to their electrical conducting nature that can interfere with RF components and their limited
integration ability with other-non-metal materials due to their wetting properties. Additionally,
for miniaturization of devices, the TIMs utilized have to shrink in size and weight
correspondingly as well. Thus only the use of very thin metal foil is allowed which decreases the
effective thermal conductivity by several orders.7 Also, the use of metallic TIMs require high
temperatures for reflow in order to achieve good surface conformity, which can range between
90°C to 450°C limiting its usage.8 For electrically insulating TIMs, typical materials include
greases, phase change materials, gels, adhesives and polymers.4, 9, 10 These TIMs are able to
achieve good surface conformity at low temperature and are relatively easy to apply.
Nevertheless, their thermal conductivities remain in the lower regime (0.1 – 0.3 Wm-1K-1).4
Many have improved these materials by infiltrating with high thermally conducting fillers, such
as metals (i.e. Ag, Au, Ni, Cu)11 and ceramics (i.e. Al2O3, AlN, BN).12, 13 This strategy increases
the thermal conductivity of commercial available TIMs to 5 – 10 Wm-1K-1.14 Despite this, this is
still much lower than typical metallic TIMs and electronics materials, which creates a bottleneck
in the thermal extraction channels (i.e. typical solder materials have a thermal conductivity
ranging between 20 – 60 Wm-1K-1 at room temperature).15-17 Although one could increase the
loading fraction of the fillers to increase the thermal conductivity, large filling ratios would also
affect its viscosity and other parameters defining the applicability, such as curing temperature,
dry-out and pump-out.
Today, the demand in performance for modern ICs has already pushed the current TIMs
beyond their limit.18 As a consequence, the performance of new generation chips has to be
lowered at times to avoid over-heating19 (which means adjusting their frequency according to the
4
temperature).20 This is obviously undesirable.21 As such, it is of importance to overcome this
heat dissipation challenge22 and many have begun to explore various nanomaterials as next
generation nano-thermal interface materials.23
Recently, there has been a trend of up-scaling two-dimensional (2D) materials, such as
graphene and h-BN, into freestanding 3D space.24, 25 The obtained structures are of ultra-light
weight (99.6% porosity, density of 1 – 5 mg cm-3), and have enhanced mechanical and surface
area properties while keeping the well-reported 2D-structural properties.26, 27 For the case of
three-dimensional graphene (3D-C), studies have shown that it has comparatively high thermal
conductivity, electrical conductivity, chemical/thermal stability and EMI shielding properties;28-
31 and for three-dimensional boron nitride (3D-BN), it has an equally high thermal conductivity
but is an electrical insulator.30 In addition, it has been reported that the interconnected nature
within the 3D-foams could drastically overcome the internal contact thermal resistance, since
single graphene domains within the 3D-foam form a continuous graphitic structure through
covalent bonding.28 It has been shown that such foam-like structures have superior thermal
interfacial characteristics to other surfaces due to the increased contact surface area.32
Furthermore, it has also been predicted that the current thermal conductivities can be further
increased by increasing the density of the 3D-foams.28 Therefore, it is believed that as the density
of these foam-like materials increases, the overall thermal extraction performance would far
exceed the thermal performance of the current TIMs.
Herein, we study the application of 3D-C and 3D-BN as an ideal electrical conductive and
insulative TIM candidate, respectively, to overcome current thermal management challenges. For
both of these high-density 3D-C and 3D-BN, characteristics that are critical to TIMs have been
investigated, which include cross-plane thermal conductivities as high as 62 – 86 Wm-1K-1,
5
interface, thermal stability upwards of 700°C and, most importantly, evaluation of the thermal
extraction efficiency on a state-of-the-art 2.5D electronic platform. Additionally, all results have
been verified with commercial software COMSOL Multiphysics simulations.
Our results have shown an exceptional performance as TIMs with a drastic decrease of chip
temperature on the 2.5D test platform from ca. 95°C to 51°C (ΔT° of 44°C) for 3D-C and to
71°C (ΔT° of 24°C) for 3D-BN at an applied power of ~5 W, which is 20% colder than any of
the commercially available TIMs tested on the same platform (i.e. Sn/Au) and among the highest
temperature decrease of hot spots on actual chips reported so far (e.g. highest values for
alternative heat spreaders currently under research range around ΔT ~ 13°C for a CVD-graphene
heat spreader,26 ΔT ~ 6 – 12°C for a graphene-based film with added silane-functionalized
molecules,27 and ΔT ~ 20°C for exfoliated few-layer graphene).33 This is a significant decrease,
since it is known that the decrease of hot spot temperature on chips by 20°C extends the
transistors lifetime by one order of magnitude.33 It must be also noted that for most alternative
materials currently under research, the cooling down of the hot spot is via the distribution of the
heat throughout the film in the in-plane direction and not by extracting it upwards (to a heat
sink). At equilibrium, such spreading of the heat would keep the entire chip at elevated
temperatures and defeats the purpose of maintaining the chip below a safe operating region.15
Contrastingly, for 3D-foams, their outstanding thermal conductivity in the cross-plane direction
and their thin thickness (which will fasten up the heat transfer upwards, rather than lateral),
makes them thus ideal candidates for solving current thermal management needs.
RESULTS AND DISCUSSION
6
A. 3D-foam fabrication and characterization
3D-foams were prepared with chemical vapor deposition (CVD) method on Ni template, as
reported previously,30 and the resulting structures are shown in Figure 1. Figure 1a and b show
optical images of the pristine 3D-C and 3D-BN structures, respectively; Figure 1c shows a
scanning electron microscopy (SEM) image of the cross-section of the 3D-foam. It can be seen
that the pores have a mean diameter of 100 – 200 µm, homogeneously distributed within the 3D-
foam. The overall thickness of the 3D-foam is of ca. 1.7 mm with a porosity of 99.6%
(corresponding to only 0.4% actual material, the rest being free space) and density of 1 mg cm-3
for the pristine 3D-BN and 5 mg cm-3 for pristine 3D-C. Figure 1d shows a transmission electron
microscope (TEM) image taken at the edge of a 3D-foam pore. It can be seen that the structure
of the 3D-foam comprises of stacked layers of graphene or h-BN, for the case of 3D-C and 3D-
BN, respectively. Typical 3D-foam has between 8 and 10 layers. The number of layers can be
tuned by the growth parameters; above 10 layers, the 3D-foam starts to resemble graphitic or
bulk h-BN structures. To increase the density of the 3D-foams, two approaches have been used;
namely, (1) prolong the growth duration and (2) by simply compressing the 3D-foam with an
applied force, as depicted in the schematic of Figure 1e. The difference between the two is
mainly that in (1), the thickness of the walls increases while the pore distribution remains
unchanged. As for (2), it decreases the pore size but the thickness of the walls remains
unchanged. To verify the quality of both 3D-foams, Raman spectroscopy was performed on both
samples. The results show a stable quality throughout the varieties of 3D-foams. The typical
peaks for a high quality multi-layer graphene, together with their ratio, are well preserved in all
three graphs (G-peak at 1580 cm-1, 2D-peak at 2700 cm-1, absence of D-peak at 1350 cm-1 and a
ratio of G:2D-peak intensity of 0.25).34
7
The electrical conductivity of the 3D-foams was measured with Van-der-Pauw method, 3D-BN
has an electrical resistivity of 16.0 x 106 Ωcm and 3D-C has an electrical resistivity of 1.7 Ωcm.
Figure 1. Material characterization of the 3D-foams. (a) Optical image of 3D-C; (b) optical
image of 3D-BN; (c) SEM image of the cross-section of a typical uncompressed 3D-foam; (d)
TEM image of the edge of a 3D-foam to determine the amount of layers comprising the 3D-
foam; (e) Raman spectroscopy on the 3D-foams with different density obtained through the two
different methods to increase layer numbers.
B. Thermal conductivity
As reported previously, the biggest constrain for the effective thermal conductivity of the 3D-
foams is the very low volume fraction (0.4%),28 hence, if the density of the 3D-foam could
increase, the thermal conductivity will rise by several orders.30
To investigate the correlation between the density of the 3D-foams and their respective thermal
conductivities, the cross-plane thermal conductivity was measured using the laser flash method35
8
since the z-plane heat spread is more crucial for the effectiveness of heat extraction for TIMs.15,
36 Supporting experiments to proof the suitability of the laser flash technique for the porous 3D-
foams is provided in the Supporting Information (Figure S1). Figure 2a and 2b show SEM
images of two different 3D-foams with different density and it can be clearly observed that the
size of the pores decreases accordingly. Figure 2a shows a sample 90 times denser than the
standard sample. The branches and pores can still be clearly distinguished, while the spacing
between individual pores reduces. Figure 2b shows the maximum density achievable, whereby
all pores have been closed and only the outline of the walls are still visible.
Obtained thermal conductivity results are shown in Figure 2c, in which the thermal conductivity
(red curve) was plotted along with its corresponding density (blue curve) for each measured
sample. Closed squares represent results for 3D-C and open squares for 3D-BN. The base
thermal conductivities of the pristine 3D-foams are of 1.2 Wm-1K-1 and 0.84 Wm-1K-1, for 3D-C
and 3D-BN respectively. As the density increases, the highest thermal conductivity values
achieved are at 86 ± 10 and 62 ± 10 Wm-1K-1 for 3D-C and 3D-BN (error deviation due to the
non-perfect surface flatness), respectively. It must be noted that 3D-BN has a slightly lower
thermal conductivity than 3D-C, due to the cumulative effect of the following structural
differences: (1) h-BN and graphene have a different phonon dispersion, especially in their
flexural modes (which in turn leads to increased phonon-phonon scattering),37-39 (2) higher
isotopic impurity in h-BN (i.e. BN has a larger isotope mixture of 19.9% 10B and 80.1% 11B than
graphite with 98.9% 12C and 1.1% 13C, which leads to an increase in phonon-isotope scattering.
This in turn reduces the thermal conductivity in BN),39, 40 (3) the difference in mass of B and N
(since the phonon frequency depends on the mass, and thus in BN local fluctuations in the
natural frequency appear which leads to additional phonon scattering)38, 41 and (4) intrinsic
9
phonon-phonon scattering due to lattice anharmonicity (also caused by the dual-character of BN,
as compared to graphene).39, 42
The results show that there is a proportional relation between density and thermal conductivity
for both types of 3D-foam. In addition, it is also worth mentioning that the two methods of
obtaining high-density 3D-foam (i.e. prolong growth time and compression of the 3D-foam)
have been compared and their thermal conductivities are comparable. The reason is that both
these methods do not change the internal configuration of the 3D-foam and thus the internal
thermal contact resistance remains low.28
The cross-plane thermal conductivities for 3D-C and 3D-BN increased 14-fold and 30-fold
over typical multilayer graphene and h-BN (i.e. graphene has a thermal conductivity of 6 and h-
BN of 1.5 – 2.5 Wm-1K-1 in the z-plane).43, 44 This is due to the isotropic interconnected structure
of the 3D-foam: while in 2D graphene (or h-BN), phonon transport is anisotropic (i.e. only along
the in-plane direction, and for out-of-plane propagation between the layers only weak Van der
Waals forces are present which do not enable efficient phonon transport), in 3D-foams, the
graphene or h-BN sheets are stitched to each other covalently and follow an isotropic structure,
which is preserved even when density is increased. This way, in 3D-foams the phonon transport
can be directed into all directions in an isotropic manner.
10
Figure 2. Study of thermal conductivity as a function of density. (a) SEM image of a 3D-foam
compressed down to 90 times its original density D1, (b) SEM image of a maximum compressed
3D-foam; (c) measured density and thermal conductivity of 3D-C and 3D-BN (inset: zoom into
the first 5 measurement points).
Also, these maximum values of thermal conductivity are among the highest cross-plane
conductivities of free-standing graphene or h-BN structures. For comparison, the closest
structure to the 3D-foams studied here is graphene and h-BN paper. For this type of structures,
thermal conductivity in the out-of-plane direction is difficult to improve further due to the
anisotropic behavior and alignment of the graphene or h-BN particles (maximum values reported
are of 1 – 5 Wm-1K-1).45 This becomes very evident when observing their cross-sectional SEM
11
images, whereby the layered structure can be clearly seen.46, 47 Further comparison to other
structures is given in Table 1.
Table 1. Comparison of cross-plane thermal conductivities among graphene and h-BN structures
Material Cross-plane
thermal conductivity
[W m-1K-1]
Remarks Ref.
Graphene 6 Weak Van der Waals forces limit conductivity between layers 44
h-BN 1.5 – 2.5 Same as graphene 43
Graphene paper max. 1 – 5 anisotropic behaviour and alignment of the G particles, which
becomes evident in the layered structure under SEM
45, 46
h-BN paper/BN-
polymer
nanocomposites
n.a.
(only in-plane)
Similar to G paper, h-BN paper is extremely layered
Pure BN has weak strength and is difficult to obtain as
freestanding film, thus needs support from another material
47-50
Commercial Silver
epoxy TIM
1.76 (pristine)
9.9 (5 vol% graphene filler)
stable in working temp range of epoxy, which is only 5 – 75 °C 16
Graphene
nanocomposite
epoxy
5.1 (10 vol%) Stable only at 5 – 75 °C 36
Graphene laminate
on PET 40 – 90 Strongly dependent on flake size
due to the amount of single flakes required not freestanding
51
Multilayer graphene
in epoxy
5 Temperature dependence of values, not freestanding 52
3D BN nanosheet
networks in epoxy
2.4 (9.29 vol%) Infiltrated in epoxy
anisotropic
53
h-BN in polyimide 7 (60 vol%) Not freestanding and requires high volume fractions 54
3D-foam 62 ± 10 (BN)
86 ± 10 (C)
Freestanding and compressible, stable up to 700 – 900 °C This
work
Also, for actual application, it is challenging to obtain freestanding graphene or h-BN
structures that can be applied directly without the need of supportive materials (such as
polymers), which further hinders the phonon transport. For 3D-foams, besides their outstanding
thermal conductivity, these 3D-foams do not require additional mixing with supporting materials
and are fully freestanding.
12
C. Interface characteristics
For TIMs, besides thermal conductivity, the quality of the interface with the chip and sink for
heat dissipation plays a major role in the heat transfer. Some of the important considerations
include: (1) thermal properties of the interface, (2) optimization of the interface, since due to the
very bad conduction of heat in air (thermal conductivity of 0.026 Wm-1K-1 at room temperature)
any presence of free space void in the joint area would increase the thermal resistance
significantly, (3) surface roughness of the contact, (4) any possible stress occurring due to
coefficient of thermal expansion (CTE) mismatches.55 To investigate these interface properties,
measurements according to ASTM5470 were performed.56-60 The value obtained is of Rth = 0.197
KW-1 (0.262 in2KW-1), which is considered a very low thermal contact resistance. This is due to
the very high surface conformity of the 3D-foam and shows that the 3D-foam is able to decrease
the interface resistance of two mating surfaces. This value is in a similar range as in reported
previously in a comparable test where 3D-C foam was placed between two Cu blocks (between
Si and Al) and pressed down, which gave a thermal resistance of 0.45 KW-1 (0.26 KW-1).32 For
comparison, the thermal interface resistance of commercial grease is around 6 times higher, at
2.71 KW-1, and silicone based adhesive is 3 times higher, at 1.35 KW-1.
To observe the morphology of compressed 3D-foam, Focused Ion Beam (FIB) and SEM have
been used to measure the thickness of the branches of the 3D-foam and the 3D-foam itself after a
compression step between two Si wafers. For this, a window has been cut through the
compressed 3D-foam in order to access the 3D-foam within the sandwiched structure (Figure
3a). It can be seen that on average, each branch is between 1.1 and 1.3 μm thick (Figure 3b).
Since the porosity of the Ni template used is between 100 and 110 ppi, and with an initial
thickness of 1.67 mm of the 3D-foam, the number of vertical pores inside the structure is
13
between 6 and 7. Since samples comprised of 10 layers are used for this application, which are
supposed to be non-compressible, then the minimum thickness the 3D-foam can reach is 6.6 μm.
The SEM image also reveals that there is still free space present between the branches. This is
due to the fact that there was a gap between the Si wafers due to its warpage,61, 62 which is larger
than the minimum thickness of the 3D-foam, thus the 3D-foam does not require full
compressibility in order to fill this gap.
In order to quantify the compressibility of the 3D-foams, a study of the relationship between
applied force and compression level is shown in Figure S2.
To further demonstrate that 3D-foam would fit into most surface roughness to improve the
thermal interface, various artificial trenches of different aspect ratio were fabricated on a Si
wafer as a test platform for extreme cases and compressed with the 3D-foam. The cross-section
of the Si–3D-foam configuration was observed under SEM and Figure 3c-f show two examples
of the trench-arrays before and after compression with the 3D-foam. Figure 3c,d show the case
of 20 µm deep, 2 µm wide trenches; Figure 3e,f of the 40 µm deep and 25 µm wide section. For
both cases, it can be seen clearly that the 3D-foam is able to completely fill up all the gaps. The
good fitting of the 3D-foam into all the trenches is due to the fact that the structure of the 3D-
foams is very refined and it is several orders smaller than the usual surface roughness or
unevenness/warpage.
14
Figure 3. SEM studies of the 3D-foam after compression on varying surface roughness. (a) FIB
cut has been performed to measure the thickness of the graphite inside the 3D-foam; (b) zoom
into the cut with measurement; (c) cross-sectional view of the 20 µm deep and 2 µm wide trench
before compression with 3D-foam and (d) after compression with 3D-foam; (e) cross-sectional
view of the 40 µm deep and 25 µm wide trench before compression with 3D-foam and (f) after
compression with 3D-foam.
D. Thermal stability
Thermal stability of the 3D-foams throughout a temperature range was assessed through
thermogravimetric analysis (TGA). This method investigates the stability of the material at
15
elevated temperatures by heating them up to ~1100°C while monitoring their weight.63, 64
Obtained curves are shown in Figure 4a, whereby the blue curve corresponds to 3D-C, which
shows that 3D-C remains unchanged up to 700°C and that only after this point, it loses 95% of
its weight; 3D-BN (green curve) on the other side remains stable up to 900°C, and then increases
in weight, which is due to the oxidation of BN to B2O3 which is heavier in mass. The initial
slight decrease of the 3D-BN mass is due to the loss of hygroscopic water.65, 66 Both results are in
good agreement with literature.67, 68 It is worth mentioning that typical operating temperatures for
electronics is of 65 – 85°C for commercial applications, 110 – 120°C for military electronic
systems, and as high as 175°C for automotive ICs and the upper spectrum of military
applications (all demarcated in the graph).69 Evidently, both 3D-foams can withstand all the
important operating temperature ranges. To ensure that the thermal conductivity of the 3D-foams
remain stable throughout the temperature range of electronics, laser flash on uncompressed 3D-C
was performed from room temperature to 200°C and Figure S3 shows the obtained results. It
can be seen that the thermal conductivity remains stable at ± 10%, which corresponds to
previously reported values for this type of 3D-foam.28, 30
Besides TGA, X-ray Photoelectron Spectroscopy (XPS) was used to analyze the thermal
stability of the chemical structure. XPS analysis was carried out on the 3D-foams after exposure
to different temperatures in air for one hour from 200°C and up to 720 °C for 3D-C and 920°C
for 3D-BN. The results are shown in Figure 4b-g. For the case of 3D-C, in the range of room
temperature to 200°C, only pure graphitic structure was detected, comprising of only carbon
aromatic groups. Starting at 200°C, an oxidation of 5 – 6% is visible, which only slightly
increases at 700°C (i.e. the point of mass loss). For the case of BN, a similar behavior can be
observed; BN oxidizes slightly with increasing temperature, but the level of oxidization remains
16
in a low regime of 8 – 17%. This very high thermal stability beyond the usual operating
temperatures is of great advantage, especially for electronics, where the increase in both
compactness and performance has generated thermal problems. In particular for electrically
insulative TIMs, one of the current challenges is their thermal stability beyond temperatures of
120 – 200°C (i.e. polymers degrade under such high temperatures,70 grease hardens).1 3D-foams
(i.e. 3D-BN for electrically insulative TIM) on the other hand are capable to withstand the
increasing demands for electronics in high temperature environment.
Figure 4. Thermal stability of the 3D-foams. (a) Measured via TGA curves of 3D-C (blue) and
3D-BN (green); and through (b, c, d) C1s XPS high resolution spectra of 3D-C at different
17
temperature ranges, (e, f, g) B1s XPS high resolution spectra of 3D-BN at different temperature
ranges.
E. Performance test in a 2.5D IC platform
To assess the heat extraction performance of 3D-foams on electronic devices, evaluation was
carried out on a 2.5D electronic test platform.71 2.5D packages refer to the packaging of multiple
dies on a Si interposer with Cu-filled Through Si Vias (TSVs) as electrical interconnects.
Dissimilar dies with different functions can be assembled together to fulfill faster operation
through shortened interconnects. Nevertheless, even though the shortened interconnect lengths
reduce the power consumption, the power dissipation for each die tends to remain unchanged.
This results in unfavorable heat density within the packages. Thus, thermal management is a key
challenge to be addressed in 2.5D packages and system designs,72, 73 whereby the TIMs
performance can play a key role in 2.5D packages cooling.
Figure 5a shows a photograph of the setup used, in which the 3D-foam is compressed directly
between the test chip and the heat sink (marked here with an arrow). Also for this setup, the 3D-
foams were first placed in their uncompressed state onto the chip and then compressed down via
the weight of the heat sink. Two types of test chip where used, one where the heating element is
exposed (denominated as “no overmold”, Figure 5b) and another where the heating source is
covered with a mold (denominated as “with overmold”, Figure 5c). The difference between the
two is that in the former case, the performance only depends on the material itself (thus a direct
evaluation of the heat spreading capabilities of the material), whereby for the latter, it is a more
application-related situation where the performance is also affected by the low conducting
overmold material, since the heat has to first go through it before being transferred to the TIM
and then the heat sink. As such, measurement with overmold would result in an overall reduction
18
of the heat extraction performance. A more detailed description of the test chip can be found in
the Experimental Section and a schematic and photograph of the system without any overmold
are shown in Figure S4. In brief, a thermal test die is placed on a Si interposer, together with two
dummy dies, which represent the logic and memory chips in a typical 2.5D integration. The
voltage of the thermal test die is gradually ramped up, while an internal diode voltage (which
detects the temperature increase in the chip) is recorded. With this, the temperature at the chip’s
junction to ambient can be deduced.
Besides the 3D-foams, other standard TIMs were evaluated in order to benchmark the
performance. Figure 5d shows the obtained results, 3D-BN is being compared directly with other
electrically non-conducting TIMs and 3D-C with other electrically conducting TIMs.
The graphs show the increase in temperature at the junction of the chip to the ambient as a
function of applied power. All materials tested have a linear relation. Only the gradients change
between the materials.
Figure 5. Test on field. (a) 2.5D test setup used; (b, c) top view of the chip used, with and
without overmold on the heater section; (d) obtained results, grouped according to device used
and electrical conductivity characteristic, mean error deviation ±0.19˚C.
19
Comparison of the curves shows that the 3D-foams perform with higher efficiency: the
measured change in temperature of the chip shows that 3D-foams are able to maintain the device
at a lower operating temperature. The slopes of the graphs are enlisted in Table 3 together with
the achieved improvement over the standard TIMs tested. It can be seen that in all these cases,
the 3D-foams achieve a significant decrease in its temperature differential; for instance, 3D-BN
is able to maintain the chip 10% and 25% cooler than its commercial electrically insulating
counterparts with and without overmold, respectively (this is equivalent to 15.06°C and 23.6°C
less heat at the point of maximum power applied during the test), and 3D-C achieves 12% and
42% improvement over its electrically conducting counterparts, equivalent to 12.56°C and
39.85°C less heat at 4.4W, for the overmold and non-overmold case, respectively.
Table 3. Temperature increase slopes for various materials tested with the 2.5D platform
Material Slope
[oC W-1]
Improvement
[%]
N
o o
verm
old
Ele
ctr
ically
insula
tin
g 3D-BN 16.6 -
Silicone TIM 22 25
Ceramic filled elastomer 22 25
Ele
ctr
ically
conductin
g 3D-C 12 -
Cu-foil 16.2 26
Sn/Au-foil 20.8 42
W
ith o
verm
old
Ele
ctr
ically
insula
tin
g 3D-BN 17.8 -
Silicone TIM 19.6 10
Ceramic filled elastomer 20.6 14
Ele
ctr
ically
conductin
g 3D-C 16.6 -
Cu-foil 18.8 12
Sn/Au-foil 18.8 12
20
Table 4 compares the obtained results with values reported by other thermal management
materials tested on different test-platforms from literature. It must be noted that the reported
values in these examples were obtained without the use of overmold on top of the hot spot. It can
be seen that for such cases, the 3D-foams outperform other materials, paired with better
applicability (e.g. one of the alternative approaches which achieved the highest temperature
decrease of 20°C, using exfoliated graphene quilts,33 is still 4°C lower than 3D-BN and 24°C
lower than 3D-C, and it requires a precise control over layer number, exfoliated size and
location, which is difficult to achieve).
Table 4. Comparison of hot spot temperature decrease of different materials
Material Achieved temperature decrease
of hot spot
Remarks Test platform Ref.
CVD-Graphene heat
spreader
ΔT~ 13°C
at 0.0156 to 0.624 W
Application of graphene not very
practical (requires PMMA transfer
to platform, then hot acetone to
remove PMMA)
Pt micro-heater, temperature
determined from measured
resistance
26
Graphene paper ΔT ~ 24-18°C
(from thermal camera image)
Orientation of films only parallel to
platform
Samples suspended above
an iron head, temperature
recorded through IR camera
46
Exfoliated graphene
quilts (few-layer
graphene)
ΔT = 20°Cb) Layer number, exfoliated size and
location are difficult to control
through exfoliation
AlGaN/GaN HFETs, ∆T
determined from
temperature-dependent
shifts in Raman peak
positions
33
Layered h-BN film ΔT = 20°C Not freestanding, requires acetate
cellulose support, polycrystalline
BN
Micro-heater configuration,
temperature recorded
through IR camera
74
Graphene-based film
with addition of silane-
functionalized
molecules
ΔT ~ 6-12°Cb)
at 2.52 W
Material only oriented in parallel Pt micro-heater, temperature
determined from measured
resistance
27
3D-foam ΔT ~ 15°Ca)b)/24°Cb) (3D-BN)
ΔT ~ 23°Ca)b)/40°Cb) (3D-C)
at 4.4 W
Orientation of structure isotropic,
improved thermal conductivity out
of plane, thus heat is extracted,
rather than spread along the film
Thermal test die,
temperature determined
through internal diode
voltage
This
work
a)Use of overmold on top of the hot-spot b)Use of heat sink on top of the hot-spot
21
Also, similarly to the 3-layer laser flash measurement, the compression with the heat sinks is
sufficient to compress down the 3D-foam to conform to the surface profile, but it does not yield
maximum compression. Nevertheless, this is sufficient enough to obtain enhancement of heat
extraction.
Besides showing the outstanding performance of the 3D-foams, these results also highlight that
heat-spreading performance not only depends on the thermal conductivity of the material used,
but also on the surface conformity. For example, the metal foil’s performance was comparable to
the ceramic filled elastomer, even though their thermal conductivities differ by more than 50
Wm-1K-1 (i.e. the thermal conductivity of the elastomer is of 3.3 Wm-1K-1, Sn/Au foil is of 57
Wm-1K-1).
Moreover, since the purpose of the TIM is to displace any micro and/or macroscopic free space
voids in between interfaces to improve heat conduction, any inclusions beyond the size of these
gaps would be redundant and eventually lower the overall interface thermal conductivity. To
further analyze the correlation between surface roughness and the optimized thickness for TIMs,
tests with 3D-foams of varying thickness were performed on the mold and heat sink with
different surface roughness and the results are summarized in Figure 6.
22
Figure 6. Delta surface roughness. (a, b, c) surface roughness profile of the three heat sinks
used; (d, e, f) corresponding results obtained using the test setup from Figure 5, mean error
deviation ±0.21˚C. (g) Experimental results for different surface roughness measured at 4 W as a
function of starting thickness of the 3D-foam; (h) theoretical results for different surface
roughness and trend of the temperature increase measured at 4 W; (i) schematics of the use of
3D-foams as TIM with different starting thickness.
The same setup of Figure 5 was used, with heat sinks of different surface roughness, namely
R1, R2 and R3 being applied. A total of three heat sinks (the surface profile of each heat sink is
shown in Figure 6a-c) were tested and for each heat sink, 4 different initial thicknesses of 3D-C
were tested (i.e. initial thickness of 1 mm, 2 mm, 5 mm and 10 mm before being compressed and
conformed to the interface)
Figure 6d-f shows the obtained graphs of temperature increase and Figure 6g shows the
obtained results plotted as a function of initial thickness of the 3D-foam. Interestingly, it can be
23
seen that an initial thickness of 2 mm is enough for all cases to achieve an effective heat
extraction (i.e. in the most roughened heat sink, the 2 mm result is nearly identical with the 5 mm
and 10 mm result).
This can be explained as follows: as mentioned previously, the 3D-foam will fill all free space
gaps between the mold and the heat sink. The resistance of the interface is thus the bulk
resistance of the 3D-foam R3D given by:
𝑅3𝐷 =𝑡𝑒
𝑘3𝐷(𝑡𝑒) (1)
where te is the effective thickness of the 3D-foam between the heat sink and the mold, k3D is the
thermal conductivity of the 3D-foam which is also dependent on the effective thickness. The
Bruggeman assumption gives the relation between the thermal conductivity and the porosity,
which also depends on the thickness.75
𝑘3𝐷 = 𝑘𝐺(1 − 𝑓(𝑡𝑒))3/2 (2)
with f the porosity and kG the thermal conductivity of bulk 3D-foam (i.e. 86 Wm-1K-1 according
to the value obtained previously at maximum compression). When the heat sink compresses the
3D-foam on the mold, it fills all the free space gaps between the two mating surfaces; however,
unlike the previous case with Si–Si, the mold is soft and will compensate for the non-flatness of
the heat sink, which means that the thermal interface resistance is only affected by the surface
roughness and the minimum thickness of the 3D-foam tmin. Due to this, it is possible to correlate
the impact of the 3D-foam’s effective thickness with the behavior of the thermal resistance for a
given surface roughness: by combining relations (1) and (2), the calculation of the thermal
resistance of the interface is computed in Figure 6h according to the initial 3D-foam thickness
for 5 different roughness tr (4, 8, 10, 15, 20 µm). The values of the effective thickness te = tmin +
24
tr (addition of the minimum thickness reachable by the film with the surface roughness) used are
summarized in Table S1.
With this relation, it can be observed that for very thin thickness the thermal resistance is very
high, which then continues to decrease until an optimal thickness is reached. This optimal
thickness point is schematized in Figure 6i top and Figure S5, whereby the 3D-foam nicely fills
in the free space gaps without creating any additional spacing between the heat sink and the mold
(thickness/gap width δ ≈ 0). This optimal value is achieved with an initial thickness of 2 mm for
the three cases of surface roughness tested here. Beyond this optimal thickness, compression of
thicker 3D-foams leads to two simultaneous effects, namely the compression of material at the
edge of the grooves (which leads to the addition of extra-material between the two mating
surfaces, depicted in Figure 6i bottom, δ >> 0) and to a higher compression level of the 3D-foam
inside the gap (as can be seen through comparison of the SEM images in Figure S5). Since a
higher compression level inside the grooves leads to lowered thermal interface resistance, but at
the same time the addition of material between the surfaces increases it, the two opposing effects
sum up to an overall less linear increase of thermal interface resistance. This becomes evident in
the case for R2, where the 5 mm 3D-foam achieves a similar performance as the 2 mm 3D-foam,
since the slight addition of extra material between the surfaces is balanced off with an increased
level of compression inside the gaps. The same applies to the case for R3 sample with the 10 mm
thick 3D-foam. This is why in Figure 6g and 6h, the 2 mm 3D-foam achieves optimal thermal
interface resistance results up to a surface roughness of 10 um and slightly thicker 3D-foams do
not change this performance. It must be noted that experimental and theoretical values match,
and the optimal thickness for each surface roughness can be reliably predicted.
25
CONCLUSION
In this work, high-density 3D-foams were studied for use as high-performance TIMs. Results
have shown that high thermal conductivities across the plane direction in the range of 62 – 86
Wm-1K-1 were achieved, which is a 14- and 30-fold increase towards their 2D counterparts, 300-
800 times higher than standard TIMs and a 10-fold increase over recent reported TIMs that are
comprised of other nanomaterials (such as graphene laminate, graphene and h-BN paper). In
addition, due to the compressible nature of the 3D-foams, a superior surface conformity was
revealed. It has also been shown that the materials can withstand temperatures up to 700 – 900°C
without any chemical changes, which is at least 500°C above current available TIMs. Direct
comparison of the 3D-foams to other state-of-the-art TIMs has shown improved cooling
performance by 20 – 30%, and comparison to other results from literature also revealed the
superior qualities of the 3D-foams. In addition, for optimum performance, it has also been shown
that for TIMs, it is important to match its thickness to the size of the free space void in between
as any thickness beyond this limit would eventually increase the thermal interface resistance.
This optimization has been demonstrated for the 3D-foams through combined simulation and
experimental results. These overall enhancements in thermal performance are due to the isotropic
thermal behavior given by the initial foam-like, interconnected structure, which is contrasting to
usual nano-TIMs, where only weak Van der Waals forces ensure the contact between layers, and
thus hindering good phonon transport.
Besides its high cross-plane thermal conductivity and excellent surface conformity, the herein
presented 3D-foams are of a free-standing material that can be applied directly at room
temperature without any need for reflow or curing. Together with their availability in both
electrically conducting and insulating mode, these 3D-foams combine the advantages of both
26
metallic TIMs (i.e. high thermal conductivity) and polymeric/adhesive-based TIMs (i.e. good
surface conformity). These advantages would allow electronics to be driven towards a higher
power regime and enable a closed-form solution15 for future electronics device package.
EXPERIMENTAL
Preparation of 3D-foams. “Growth of 3D-BN and 3D-C was carried out in a split tube
furnace using Ni foam (Latech Scientific Supply Pte. Ltd.) as a catalytic substrate. For BN, 0.5 g
of ammonia borane (NH3-BH3) is placed in a ceramic boat away from the heating element in the
quartz tube. The Ni foam is annealed for 2 hours at 1000°C at low pressure (50 Pa), under H2
flow, and is subsequently followed by BN growth which is initiated by heating the ammonia
borane at 120°C, and the pressure is kept constant at 60 Pa. The growth for 3D-C was carried out
similarly, with CH4 as the C precursor gas (and shorter annealing of 5 minutes only). After
growth, the lid of the furnace is lifted for fast cooling. The as-grown 3D-BN/C/Ni sample is dip
coated with poly(methyl methacrylate) (PMMA) to protect the BN/C layers and then immersed
into hot diluted hydrochloric acid (HCl) for at least 5 hours until the Ni is completely etched
away. Finally, to remove the PMMA coating, the samples are annealed at 700 °C for 1 hour (in
air for the case of 3D-BN and in Ar and H2 environment for 3D-C).”30
Measurements. “SEM images were recorded with a JEOL JSM 5600LV at 15 keV, Raman
spectroscopy (WITec CRM200 Raman with an Nd:YAG 532 nm laser as excitation source) was
performed at room temperature to determine the crystalline configuration of the films. Thermal
conductivity measurements were performed using the Laser Flash technique (LINSEIS XFA
500) at room temperature by first measuring thermal diffusivity α (mm2s-1).76 Based on these
results, the thermal conductivity κ (Wm-1K-1) can be calculated through Formula (3):
27
)(
)()(
Tc
TT
P (3)
where cp (kJ kg-1K-1) is specific heat and ρ (kg m-3) reference density at room temperature.”31
The error deviation was calculated by measuring 5 different spot locations on each sample. To
ensure the correctness and repeatability of the last (fully compressed) sample, another set of 5
measurements was carried out on a second sample (i.e. a total of 10 measurements on 2 different
samples for the fully compressed case). Specific heat was measured prior to the laser flash
measurement using differential scanning calorimetry (DSC Q100, TA Instruments). The mass
density was obtained by weighing a sample of known dimensions with a precision electronic
balance (Mettler Toledo). The thermal conductivity is in cross-plane, meaning that it goes
through the thickness-direction of the film. Electrical conductivity was performed using the Van-
der-Pauw method on a four-probe Karl SUSS 200 Micro-Prober. Long term stability
measurements were performed via TGA (DTG-60H, in air environment at 50 ml min-1) and XPS
(Thermo scientific theta probe XPS). Measurements of thermal contact resistance are conducted
according to ASTM 5470 using a commercial system (AnalysisTech TIM Tester model 1400).
The principle of this method is based on imposing a one-dimensional heat flow across the sample
and measure the resulting temperature difference. The sample thermal resistance, Rth, is then
defined as the ratio of the temperature difference to the heat flow. According to the Standard, the
3D-foams are considered Type II materials (low contact resistances, elastic and plastic
deformations combined with an elasticity increase with deformation), so a controlled contact
pressure is applied. Sample size is of 33 mm diameter disks, test temperature is of 50°C. Used
pressure is of 10 psi, which yields full compression.
28
Test on field. “A thermal test die (5.08 mm × 5.08 mm) consisting of four heating unit cells,
covering over 85% of the chip area to provide uniform heating, and six diodes for die
temperature sensing, was fabricated onto a Si interposer (18 mm × 18 mm × 0.1 mm) together
with two dummy dies (7.6 mm × 10.9 mm and 8 mm × 8 mm), which represent the logic and
memory chips in a typical TSI integration. All the dies were fabricated and assembled on the
same interposer wafer through microbumps and underfill processes. The interposer was then
assembled onto an organic substrate (31 mm × 31 mm × 1 mm). For the molded packages,
additional molding process was conducted on the wafer level to form the epoxy molded
encapsulation on the chip package. A D-type edge connector was used to connect the thermal test
chip to the power supply and diodes for thermal testing. The thermal diodes were calibrated in
the air-convection oven with data taken at room temperature, 60°C, 100°C, and 140°C to obtain
the proportionality factor (K factor in semiconductor thermal management). The six diodes on
the thermal test chip have excellent linearity with R-squared value ≥ 0.9999, with almost
identical K factor of 517 KV-1 ±1.2 KV-1 based on four calibrated packages on board. During the
test, a power supply from Xantrex (XHR 150-7) was used to supply electrical power to the
thermal chip, whereas the temperature reading was attained by activating the diodes with 1 mA
current from the Keithley 2400 sourcemeter. For temperature increase test, the package was first
placed in still air for 10 – 15 minutes without power input and then the diode voltage (Vdiode) and
the ambient temperature (Ta,0) were recorded as initial readings. A J-type thermocouple with
factory calibration was used to measure the ambient temperature. The power input was then
applied, and the corresponding junction temperatures and ambient temperature were recorded
after reaching thermal equilibrium. The corresponding junction Temperature TJ was then
calculated as follows:
29
0,aJ TKVT (4)
where Ta,0 is the ambient temperature before the power is input and ΔV the average of the diode
voltage difference before the power input.”71, 77
ASSOCIATED CONTENT
Supporting Information. Figures S1 – S5, Table S1. This material is available free of charge
via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
*E-mail: [email protected]
ACKNOWLEDGMENT
The authors would like to acknowledge the funding support from Singapore Ministry of
Education Academic Research Fund Tier 2 No. MOE2013-T2-2-050.
REFERENCES
1. http://www.businesswire.com/news/home/20151216005646/en/IDTechEx-Research-
More-than-Moore-electronics-packaging, retrieved 16 December 2015
2. Tong, X. C., Advanced materials for thermal management of electronic packaging.
Springer Science & Business Media: 2011; Vol. 30.
3. Viswanath, R.; Wakharkar, V.; Watwe, A.; Lebonheur, V. Intel Technology Journal Q3
2000.
4. Prasher, R. Proc. IEEE 2006, 94, 1571-1586.
5. Schulz, M., The Challenging Task of Thermal Management. In PCIM, Nürnberg 2011,
2011.
6. Prasher, R.; Chiu, C.-P., Thermal Interface Materials. In Materials for Advanced
Packaging, Lu, D.; Wong, C. P., Eds. Springer US: 2009; pp 437-458.
7. Nath, P.; Chopra, K. Thin Solid Films 1974, 20, 53-62.
8. Green, R. E., Machinery's handbook. Industrial Press: 1992; Vol. 24.
30
9. Farhad, S.; Whalley, D. C.; Conway, P. P. In Thermal Interface Materials - A Review of
the State of the Art, Electronics Systemintegration Technology Conference, 2006. 1st, 5-7 Sept.
2006, 2006; pp 1292-1302.
10. Liu, J.; Michel, B.; Rencz, M.; Tantolin, C.; Sarno, C.; Miessner, R.; Schuett, K.-V.;
Tang, X.; Demoustier, S.; Ziaei, A. In Recent progress of thermal interface material research-an
overview, Thermal Inveatigation of ICs and Systems, 2008. THERMINIC 2008. 14th
International Workshop on, 2008; IEEE: pp 156-162.
11. Li, Y.; Wong, C. P. Mater. Sci. Eng. R Rep. 2006, 51, 1-35.
12. Fu, J.; Shi, L.; Zhang, D.; Zhong, Q.; Chen, Y. Polym. Eng. Sci. 2010, 50, 1809-1819.
13. Zhu, B.; Ma, J.; Wu, J.; Yung, K.; Xie, C. J. Appl. Polym. Sci. 2010, 118, 2754-2764.
14. Huang, X.; Jiang, P.; Tanaka, T. IEEE Elect. Insul. Mag. 2011, 27, 8-16.
15. Moore, A. L.; Shi, L. Mater. Today 2014, 17, 163-174.
16. Goyal, V.; Balandin, A. A. Appl. Phys. Lett. 2012, 100, 073113.
17. King, J., Materials handbook for hybrid microelectronics. Artech House Inc: 685 Canton
Street, Norwood, Massachusetts 02062, USA, 1988.
18. Garimella, S. V.; Singhal, V.; Dong, L. Proc. IEEE 2006, 94, 1534-1548.
19. http://www.extremetech.com/extreme/222590-an-end-to-scaling-intels-next-generation-
chips-will-sacrifice-speed-to-reduce-power, retrieved 20th May 2016
20. https://www.pugetsystems.com/labs/articles/Impact-of-Temperature-on-Intel-CPU-
Performance-606/ , retrieved 20th May 2016
21. http://www.anandtech.com/show/9550/understanding-intels-dynamic-power-and-
thermal-framework-smarter-throttling, retrieved 20th May 2016
22. Garimella, S. V.; Fleischer, A. S.; Murthy, J. Y.; Keshavarzi, A.; Prasher, R.; Patel, C.;
Bhavnani, S. H.; Venkatasubramanian, R.; Mahajan, R.; Joshi, Y. IEEE Trans. Compon. Packag.
Technol. 2008, 31, 801-815.
23. Bar-Cohen, A.; Matin, K.; Narumanchi, S. J. Electron. Packag. 2015, 137, 040803.
24. Chen, Z.; Ren, W.; Gao, L.; Liu, B.; Pei, S.; Cheng, H.-M. Nat. Mater. 2011, 10, 424-
428.
25. Yin, J.; Li, X.; Zhou, J.; Guo, W. Nano Lett. 2013, 13, 3232-3236.
26. Gao, Z.; Zhang, Y.; Fu, Y.; Yuen, M. M.; Liu, J. Carbon 2013, 61, 342-348.
27. Zhang, Y.; Han, H.; Wang, N.; Zhang, P.; Fu, Y.; Murugesan, M.; Edwards, M.; Jeppson,
K.; Volz, S.; Liu, J. Adv. Funct. Mater. 2015, 25, 4430-4435.
28. Pettes, M. T.; Ji, H.; Ruoff, R. S.; Shi, L. Nano Lett. 2012, 12, 2959-2964.
29. Chen, Z.; Xu, C.; Ma, C.; Ren, W.; Cheng, H.-M. Adv. Mater. 2013, 25, 1296-1300.
30. Loeblein, M.; Tay, R. Y.; Tsang, S. H.; Ng, W. B.; Teo, E. H. T. Small 2014, 10, 2992-
2999.
31. Loeblein, M.; Bolker, A.; Tsang, S. H.; Atar, N.; Uzan-Saguy, C.; Verker, R.; Gouzman,
I.; Grossman, E.; Teo, E. H. T. Small 2015, 11, 6425-6434.
32. Zhang, X. F.; Yeung, K. K.; Gao, Z. L.; Li, J. K.; Sun, H. Y.; Xu, H. S.; Zhang, K.;
Zhang, M.; Chen, Z. B.; Yuen, M. M. F.; Yang, S. H. Carbon 2014, 66, 201-209.
33. Yan, Z.; Liu, G.; Khan, J. M.; Balandin, A. A. Nat. Commun. 2012, 3, 827.
34. Ferrari, A. C.; Meyer, J. C.; Scardaci, V.; Casiraghi, C.; Lazzeri, M.; Mauri, F.; Piscanec,
S.; Jiang, D.; Novoselov, K. S.; Roth, S.; Geim, A. K. Phys. Rev. Lett. 2006, 97, 187401.
35. Lim, K.-H.; Kim, S.-K.; Chung, M.-K. Thermochim. Acta 2009, 494, 71-79.
36. Shahil, K. M.; Balandin, A. A. Nano Lett. 2012, 12, 861-867.
31
37. Nag, A.; Raidongia, K.; Hembram, K. P. S. S.; Datta, R.; Waghmare, U. V.; Rao, C. N.
R. ACS Nano 2010, 4, 1539-1544.
38. Sevik, C.; Kinaci, A.; Haskins, J. B.; Çağın, T. Phys. Rev. B: Condens. Matter 2011, 84,
085409.
39. Lindsay, L.; Broido, D. Phys. Rev. B: Condens. Matter 2011, 84, 155421.
40. Sevik, C.; Kinaci, A.; Haskins, J. B.; Çağın, T. Phys. Rev. B: Condens. Matter 2012, 86,
075403.
41. Che, J.; Çağın, T.; Deng, W.; Goddard III, W. A. J. Chem. Phys. 2000, 113, 6888-6900.
42. Sichel, E.; Miller, R.; Abrahams, M.; Buiocchi, C. Phys. Rev. B: Condens. Matter 1976,
13, 4607.
43. Simpson, A.; Stuckes, A. J. Phys. C: Solid State Phys. 1971, 4, 1710.
44. Delhaes, P., Graphite and precursors. CRC Press: 2000; Vol. 1.
45. Wu, H.; Drzal, L. T. Carbon 2012, 50, 1135-1145.
46. Xin, G.; Sun, H.; Hu, T.; Fard, H. R.; Sun, X.; Koratkar, N.; Borca-Tasciuc, T.; Lian, J.
Adv. Mater. 2014, 26, 4521-4526.
47. Zhu, H.; Li, Y.; Fang, Z.; Xu, J.; Cao, F.; Wan, J.; Preston, C.; Yang, B.; Hu, L. ACS
Nano 2014, 8, 3606-3613.
48. Song, W.-L.; Wang, P.; Cao, L.; Anderson, A.; Meziani, M. J.; Farr, A. J.; Sun, Y.-P.
Angew. Chem. Int. Ed. 2012, 51, 6498-6501.
49. Yao, Y.; Zeng, X.; Wang, F.; Sun, R.; Xu, J.-b.; Wong, C.-P. Chem. Mater. 2016, 28,
1049-1057.
50. Meziani, M. J.; Song, W.-L.; Wang, P.; Lu, F.; Hou, Z.; Anderson, A.; Maimaiti, H.; Sun,
Y.-P. ChemPhysChem 2015, 16, 1339-1346.
51. Malekpour, H.; Chang, K. H.; Chen, J. C.; Lu, C. Y.; Nika, D. L.; Novoselov, K. S.;
Balandin, A. A. Nano Lett. 2014, 14, 5155-5161.
52. Li, Q.; Guo, Y.; Li, W.; Qiu, S.; Zhu, C.; Wei, X.; Chen, M.; Liu, C.; Liao, S.; Gong, Y.;
Mishra, A. K.; Liu, L. Chem. Mater. 2014, 26, 4459-4465.
53. Zeng, X.; Yao, Y.; Gong, Z.; Wang, F.; Sun, R.; Xu, J.; Wong, C. P. Small 2015, 11,
6205-6213.
54. Sato, K.; Horibe, H.; Shirai, T.; Hotta, Y.; Nakano, H.; Nagai, H.; Mitsuishi, K.; Watari,
K. J. Mater. Chem. 2010, 20, 2749-2752.
55. Casalegno, V.; Vavassori, P.; Valle, M.; Ferraris, M.; Salvo, M.; Pintsuk, G. J. Nucl.
Mater. 2010, 407, 83-87.
56. Xu, J.; Fisher, T. S. Int. J. Heat Mass Transfer 2006, 49, 1658-1666.
57. Jun, X.; Fisher, T. S. IEEE Trans. Compon. Packag. Technol. 2006, 29, 261-267.
58. Xu, J.; Fisher, T. S., Enhancement of thermal interface conductivities with carbon
nanotube arrays. US Patent 8,093,715: 2012.
59. Zhang, K.; Chai, Y.; Yuen, M. M. F.; Xiao, D.; Chan, P. Nanotechnology 2008, 19,
215706.
60. Culham, J.; Teertstra, P.; Savija, I.; Yovanovich, M. In Design, assembly and
commissioning of a test apparatus for characterizing thermal interface materials, Thermal and
Thermomechanical Phenomena in Electronic Systems, 2002. ITHERM 2002. The Eighth
Intersociety Conference on, 2002; IEEE: pp 128-135.
61. Leroy, B.; Plougonven, C. J. Electrochem. Soc. 1980, 127, 961-970.
32
62. Draney, N. R.; Liu, J. J.; Jiang, T. In Experimental investigation of bare silicon wafer
warp, Microelectronics and Electron Devices, 2004 IEEE Workshop on, 2004; IEEE: pp 120-
123.
63. Due, J.; Robinson, A. J. Appl. Therm. Eng. 2013, 50, 455-463.
64. Goel, N.; Anoop, T.; Bhattacharya, A.; Cervantes, J.; Mongia, R. K.; Machiroutu, S. V.;
Lin, H.-L.; Huang, Y.-C.; Fan, K.-C.; Denq, B.-L. In Technical review of characterization
methods for thermal interface materials (TIM), Thermal and Thermomechanical Phenomena in
Electronic Systems, 2008. ITHERM 2008. 11th Intersociety Conference on, 2008; IEEE: pp 248-
258.
65. Vieira, C.; da Silva, P.; da Silva, F.; Capitaneo, J.; Monteiro, S. Rev. Mater. 2005, 10,
526-536.
66. Mo, L.; Leong, K. K. M.; Kawi, S. Catal. Sci. Tech. 2014, 4, 2107-2114.
67. Pang, L. S. K.; Saxby, J. D.; Chatfield, S. P. J. Phys. Chem. 1993, 97, 6941-6942.
68. Xie, S.-Y.; Wang, W.; Fernando, K. A. S.; Wang, X.; Lin, Y.; Sun, Y.-P. Chem.
Commun. 2005, 3670-3672.
69. Tummala, R. R., Fundamentals of microsystems packaging. McGraw-Hill New York:
2001.
70. Tansley, T.; Maddison, D. J. Appl. Phys. 1991, 69, 7711-7713.
71. Zhang, H. Y.; Zhang, X. W.; Lau, B.; Lim, S.; Ding, L.; Yu, M. IEEE Trans. Compon.
Packag. Manuf. Technol. 2014, 4, 807-816.
72. Lau, J. H., TSV Interposers: The Most Cost-Effective Integrator for 3D IC Integration. In
Chip Scale Rev, 2011; Vol. 15, pp 23-27.
73. Hoe, Y. Y. G.; Yue, T. G.; Damaruganath, P.; Chong, C. T.; Lau, J. H.; Xiaowu, Z.;
Vaidyanathan, K. In Effect of TSV interposer on the thermal performance of FCBGA package,
Electronics Packaging Technology Conference, 2009. EPTC'09. 11th, 2009; IEEE: pp 778-786.
74. Sun, S.; Bao, J.; Mu, W.; Fu, Y.; Zhang, Y.; Ye, L.; Liu, J. In Cooling hot spots by
hexagonal boron nitride heat spreaders, Electronic Components and Technology Conference
(ECTC) , 2015 IEEE 65th, 26-29 May 2015, 2015; pp 1658-1663.
75. McLachlan, D. S. J. Phys. C: Solid State Phys. 1985, 18, 1891.
76. Min, S.; Blumm, J.; Lindemann, A. Thermochim. Acta 2007, 455, 46-49.
77. Heng Yun, Z.; Xiao Wu, Z.; Lau, B. L.; Lim, S.; Liang, D.; Yu, M. B. IEEE Trans.
Compon. Packag. Manuf. Technol. 2014, 4, 807-816.